NVIDIA is attracting attention by growing rapidly in sales of artificial intelligence (AI) cloud services and software beyond GPU chips.
In keeping with The Information, Nvidia has recently shown confidence to investors that it could possibly expect annual sales of $150 billion (about 218 trillion won) from its cloud and software business in the long run.
This can be a figure that far exceeds the present annual sales of Amazon Web Services (AWS), and is a story that ought to make existing cloud corporations nervous.
NVIDIA is thought to have quietly expanded its cloud services over the past two years. It leases GPU servers on to corporations and in addition provides software to assist develop AI applications.
NVIDIA CEO Jensen Huang also expressed his intention to strengthen services similar to ‘DGX Cloud’ and AI software platform ‘AI Enterprise’. “That is more likely to be a really significant business over time,” he said during his quarterly earnings conference call in February. “We expect every company on the planet to be buying software in addition to chips.”
Particularly, Nvidia CFO Colette Kress predicted that this 12 months’s sales in software, services, and support businesses will exceed $2 billion (about 3 trillion won). Even though it is small in comparison with GPU sales revenue, it shows that there’s great growth potential within the AI cloud market.
Microsoft (MS) announced that it has earned $1 billion (roughly KRW 1.5 trillion) annually from its ‘Azure Open AI’ service, and Google is targeting $1 billion in sales in North America this 12 months with its AI service. CoreWeave, which provides cloud services based on NVIDIA GPUs, expects sales of roughly $2 billion this 12 months.
Moreover, as cloud corporations similar to AWS, Microsoft, and Google Cloud are accelerating the event of AI chips to exchange GPUs, NVIDIA’s expansion of its cloud and software business is claimed to be receiving positive reviews from investors.
NVIDIA’s move could possibly be a burden on existing cloud corporations. The structure is such that NVIDIA forms a direct relationship with its customers and uses existing cloud providers as intermediaries.
Currently, NVIDIA provides DGX cloud services using the infrastructure of existing cloud corporations similar to AWS, MS, and Google, slightly than its own data center. NVIDIA pays infrastructure costs to cloud corporations after which resells it at a premium price. DGX Cloud not only helps corporations develop AI models, but additionally provides pre-trained AI models to optimize for your online business.
Meanwhile, if NVIDIA builds its own data center and provides DGX cloud services, it could possibly be a fatal blow to existing cloud corporations. It’s because NVIDIA can distribute the newest GPU chips, that are difficult to acquire immediately, to its own data centers first. Moreover, by running its own data center, NVIDIA will not must pay infrastructure costs to cloud corporations.
Ultimately, the evaluation is that if NVIDIA provides DGX cloud services directly from its own data center without utilizing existing cloud service infrastructure, it is going to compete directly with big tech cloud corporations.
Nevertheless, CEO Jensen Huang’s decision on whether to maneuver on this direction has not yet been clearly revealed.
Reporter Park Chan cpark@aitimes.com
